View
5
Download
0
Category
Preview:
Citation preview
(Multinational Finance Journal, 2004, vol. 8, no. 3 & 4, pp. 173–209)© by Multinational Finance Society, a nonprofit corporation. All rights reserved.DOI: 10.17578/8-3/4-2
1
Testing for Multiple Types of Marginal Investorin Ex-Day Pricing
Jan BartholdyAarhus School of Business, Denmark
Kate BrownUniversity of Otago, New Zealand
The observed changes in share prices at the ex-dividend day have ledresearchers to look for a single marginal investor, either a long or a short termtrader with different tax status, dominating all trades to explain the ex-daypricing in different markets. This paper provides a model which extends thisresearch in three directions. One, it allows for the possibility that different typesof traders may influence different stocks, thereby generating a separatingequilibrium. Two, it identifies an additional marginal investor who has the optionof being taxed as a short term or long term trader. Three, it explicitly models thefact that it can take a considerable time from when a dividend based trade ismade until taxes have to be paid on that trade. A unique data set from NewZealand is used for the empirical analysis. Evidence of a separating equilibriumwith at least two types of marginal investors is found (JEL: G10, G35).
Keywords: dividends, ex-day pricing, taxation.
I. Introduction
The finance literature contains extensive empirical evidence that shareprices fall by an amount less than the dividend per share on the day thestock goes ex-dividend. The search for an explanation for thisphenomenon has focused on identification of one type of marginalinvestor that dominates trades in all stocks. Three types of marginalinvestors have been suggested. The first type of marginal investor, oftenreferred to as long term investors, faces different taxes on dividends and
Multinational Finance Journal174
1. Elton and Gruber (1970), Eades et al. (1984), Brown and Walter (1986), Chui et al.(1992), Barclay (1987), Robin (1991), Michaely (1991), Hearth and Rimbey (1993),Michaely and Murgia (1995) and Bartholdy and Brown (1999).
2. Kalay (1982), Lakonishok and Vermaelen (1983), Booth and Johnston (1984) andHeath and Jarrow (1988).
3. Boyd and Jagannathan (1994).
capital gains.1 The second type of marginal investor, referred to as shortterm or professional investors, faces the same tax rate on dividends andcapital gains.2 The third type of marginal investor is corporate investorswho act as dividend captures.3 The early papers (referenced in footnotes1 and 2) all suggest that all trades are dominated by one type ofinvestors whereas Boyd and Jagannathan (1994) and Koski and Scruggs(1998) suggest the existence of a separating equilibrium with two orthree types of traders trading in different classes of shares. Koski andScruggs analyze trading around ex-days and abnormal trading volumeby security dealers and find some abnormal trading by dividend capturesbut little evidence of long term traders exposed to different tax rates.
This paper extends the existing literature on ex-day pricing in threedifferent ways. First, it extends the traditional model used to test for taxeffects in ex-day pricing to account for differences in timing betweenex-day trading and payment of associated taxes. In the traditionalmodels it is assumed that both dividend and capital gains taxes are paidon the ex-day, whereas in this model explicit account is taken fordifferences in timing. Depending on the tax system this timingdifference may be up to two years. Secondly, the model introduces anew (to the literature) marginal investor and allows for a separatingequilibrium representing three different marginal traders: professionalshort term traders who pay income tax on dividends and capital gainsand can deduct losses; long term traders who pay different dividend andcapital gains tax, and a (new) mixed trader who can choose whether tobe taxed as a short or long term trader. Finally, a novel empiricalmethodology is used to test for the existence of a separating versusdominating equilibrium and the identity of the marginal investors.
Most prior tests of the hypotheses relating to ex-dividend day pricingbehavior have only had access to data sets of either taxable dividendsor non-taxable dividends and focused on one type of marginal investor.The data set used in this paper is unique since it contains individualcompanies whose dividends were taxable or non-taxable depending onthe source of the funds supporting the dividend. This tax difference
175Ex-Dividend Pricing
allows for powerful tests of tax effects and of marginal investors. InNew Zealand, between 1982 and 1985, some companies issued bothtaxed and non-taxed dividends on a regular basis. Further, the tax rulesallowed individuals to choose their tax status with regard to capitalgains, leading to the creation of an option. Finally, short selling was notpermitted during this period. The unique combination of tax treatmentsmakes it possible to test whether different types of marginal investorsaffect the pricing of different stocks, i.e. a separating equilibrium, orone type of marginal investor dominating all stocks, i.e., a dominatingequilibrium. The results support a separating equilibrium with two typesof marginal investors: a long term investor and a mixed trader.
Bartholdy and Brown (1999) use an equivalent dataset to test for taxeffects in New Zealand but do not take into account the timing ofpayment of taxes that in some circumstances can be up to two yearsafter receiving the dividends, nor that different marginal investors mayaffect trades in different stocks.
The remainder of this paper is organised as follows. Section IIcontains a brief description of the taxation of dividends and capitalgains in New Zealand. In section III the tax regime is modelled moreformally. Section IV contains a discussion of the data used in the study.Sections V and VI test for a dominating and separating equilibriumrespectively and section VII concludes the paper.
II. Institutional Setting in New Zealand from 1982 to 1985
The New Zealand Income Tax Act of 1976 provided for both taxableand non-taxable dividends. If dividends were distributions of capitalreserves from, for example, realised profits on the sale of asset, thenthey were tax free. Otherwise, dividends were taxed as income to theindividual investor. This taxation regime lasted until 1985 after whichstrong restrictions were imposed on the issue of tax free dividends. Until1985 the same company could issue both taxable and non-taxabledividends on the same day.
Another unique feature of the New Zealand tax regime was thetaxation of capital gains. The tax-code differentiated between long term(passive) buyers with zero capital gains tax and short term (active)buyers who faced taxation on capital gains. Differentiation was basedon the intention of the purchaser at the purchase date. If the shares werebought for long term gain as an investment, there were no taxes on gains
Multinational Finance Journal176
4. As far as the authors can ascertain, New Zealand was the only country with thiscombination of tax laws.
and losses could not be deducted. If the shares were bought for thepurpose of trading in an active portfolio, any gain on the sale wastaxable and losses could be deducted.4
Under the tax code it is possible to identify three types of marginalinvestors: passive or long term, active or short term, and mixed. Longterm investors who did not actively trade could safely avoid taxation oncapital gains, but were not able to deduct losses. Short term investorswho traded all the stocks in the portfolio regularly were taxed on capitalgains and dividends, but were able to deduct losses. Mixed investorsmaintained two portfolios: one active trading portfolio where all thestocks were traded regularly and one passive portfolio were stocks werepurchased for long term gain.
Mixed traders could buy the stock cum-dividend with the option ofholding the stock for long term gain, taxed as a passive trader, or sellingit on the ex-day for short term taxable profits, taxed as a short termtrader. As long as the investor had in the past engaged in transactionsfor the purpose of short term gains, it would have been reasonably easyto argue to the tax authorities that the purpose of the trade was shortterm profit. Alternatively, if the investor decided to hold the stock inhis/her investment portfolio, it could be justified as a tax exempt longterm investment.
Thus, the tax status of the capital gains was determined by theinvestor’s ability to convince the tax department of the status of theindividual transactions. An investor’s choice depended on the relativevalues of the total after-tax expected gains. In practice, most smallprivate investors were classified as passive or long term investorswhereas institutional investors were classified as active or mixed. If theinstitutional investor maintained two sets of books; a “passive book”and an “active book,” then, when buying a stock cum-dividend, thesetraders had the option of selling the stock when it went ex-dividend (andrecord the transaction in the “active book”) and be taxed as an activetrader or holding the stock (and entering it in the “passive book”) andbe taxed as a passive trader.
During the period up to 1985 short selling was not allowed in NewZealand making it difficult for one type of investor to dominate alltrades.
For most investors in New Zealand, the tax year ended on March 30,with tax returns filed in June or July depending on sources of income.
177Ex-Dividend Pricing
5. New Zealand had a one hundred percent deductibility of intercorporate dividends inthe relevant time period.
For salary earners the most common situation was that any taxes payablewere not due until February in the following year. For tax payers withsignificant non-salary income, such as dividend and capital gain income,provisional tax based on the prior year’s tax liability/short fall was paidduring the year. As a result of the timing of the payments, an investorengaging in trading around ex-dividends may not have taxes due fornearly two years after the transaction. The current year’s provisional taxwas independent of the current investment decision. This arrangementmeant that the purchase and sale of shares in April, for example,generated tax effects that were not payable for nearly two years.
Boyd and Jagannathan (1994) suggest three classes of traders basedon different treatments under the tax code. For the U.S. they suggestedtaxable individuals, tax-advantaged dividend captures, and tax-neutralarbitrageurs. Taxable individuals are similar to passive investors andtax-neutral arbitrageurs are similar to active investors, with somequalifications. The suggestion of dividend capture, in which corporateinvestors were able to deduct all or part of the dividend income receivedfrom investment in the shares of other companies, was considered.5
However, an examination of the accounts of listed companies revealedthat the majority of the largest companies in New Zealand neverreceived dividend income from investments, suggesting that dividendcapture was not a feature of corporate strategy, at least among the listedfirms. This group of investors was not analyzed further.
For the period of this paper, 1982 to 1985, the shares were traded onthe New Zealand Stock Exchange in Auckland which was organised asan “open auction market”. The bid and ask prices were called out fromthe floor and recorded on a large black board by runners. Transactionscosts were on a sliding scale ranging from 2.5% to 1% depending on thesize of the transaction.
III. The General Model
The expected payoff at the time of purchase from investing in a stockis equal to the expected capital gain minus taxes plus dividends. Thesecash flows occur at different times so it is necessary to use presentvalues to account for the time value of money.
Multinational Finance Journal178
FIGURE 1.—Taxation Time-line for Shares with Taxable Dividends
The general situation is shown in figure 1. An investor can buy thestock with dividends on the cum-day, t0, or without dividends on theex-day, t1. The actual dividends are received Tp days later. However,taxes on the dividends are not due until Tdt days after the cum-day whichmay be up to nearly two years. These large time differences necessitatecareful attention when calculating the present values of the cash flowsto evaluate the various trading strategies. The net present value of thethree cash flows from investing in a stock; dividends and dividendtaxes, capital gains tax and the final sales price, are discussed below.
Dividends and Dividend Taxes
Dividends are not received on the ex-day, but Tp days later. The amountof the dividend has to be discounted back to time zero (cum-day):
( ) p f
p p
T rT NT T NTt tD D D D e−+ = +
where are respectively taxable and non-taxable dividendsandp p
T NTt tD D
paid at tp, rf is the risk free rate of interest, DT is the present value of thetaxable dividends and DNT is the present value of the non-taxabledividends paid at time tp.
179Ex-Dividend Pricing
Subsequent income taxes on dividends are not due until Td days afterthe purchase of the stock. As indicated above, in New Zealand this maytake up to two years. The present value of the future dividend tax is:
,dt f
dt
T rdtl Td d tD D eτ τ −=
where τd is the tax rate on dividends, Tdt is the number of days topayment of dividend taxes. The cash flows are known and thereforediscounted by the risk free rate of interest. Ddtl is therefore the presentvalue of the tax on dividends.
Capital gains tax
The capital gains tax from a sale of the stock at time ts is:
( ) ( )0 0s sc t t c t c tE P P E P Pτ τ τ⎡ ⎤− = −⎣ ⎦
The sale of shares occurs after Ts days (at time ts). For active traders itis assumed that the stock is sold on the ex-day, that is Ts = 1 whereas forpassive traders Ts > 1. Capital gains tax on this transaction is due afterTc days where Tc > Ts, and capital gains taxes may not be due untilnearly two years after the sale of the stock. The present value of thesecond term is straightforward, the amount is known at time zero andthe discount rate is therefore the risk free rate of interest. From the timeof purchase to the time of sale, the present value of the first term is astochastic variable and it therefore has to be discounted for the first Ts
days by the risk adjusted rate. After the sale of the stock until thepayment of taxes, the value is known and the risk free rate is thereforeused for the period Tc–Ts. The present value of the capital gains tax attime zero is therefore:
,( ) ( )0
s f c s f c
s
T r r T T r Tc c t tCG E P e P eτ τ − − − −⎡ ⎤= −⎣ ⎦
where r is the risk adjusted rate.
Sales Price
Finally, the expected sales price, Pts, is discounted back Ts days to timezero using the risk adjusted rate, r:
Multinational Finance Journal180
.( ) ( ) s
s
rTtE P E P e−=
Expected Payoff from Investment Strategies
The expected payoff at the cum-day, t0, from buying and holding thestock for s days is therefore given by:
, (1)( ) ( )0 0
NT T dtlt t c dE P CG D D D E Pπ τ τ= − − + + − +
where t0 is the cum-dividend date. This expression is analyzed below forthe three different kinds of marginal investors.
A. Passive or Long Term Investor
The passive trader can buy the stock on the cum-day or on the ex-dayand sell it after s days. Recall that a passive trader does not pay capitalgains tax and the expected payoff from buying on the cum-day is:
. (2)( ) ( )0 0
T NT dtlt d tE D D D E P Pπ τ= + − + −
If a passive trader buys a stock on the ex-day, t1, and holds it for “s–1”days, the expected payoff at time zero from this transaction is:
. (3)( ) ( ) ( )0 1
rt tE E P e E Pπ −= − +
For the markets to be in equilibrium, these two strategies must result inthe same expected payoff. Therefore, the equilibrium condition for apassive trader is:
, (4)( )0 1
r T NT dtlt t dP E P e D D Dτ−− = + −
which for stock i can be rewritten as:
, (5)0 ,
AD dtlt i i d tP D DτΔ = −
where:
and( )0 0 1, , ,
rt i t i t iP P E P e−Δ = −
.AD T NTi i iD D D= +
181Ex-Dividend Pricing
If discounting of various cash flows is ignored, (i.e. setting Tp= Td= 0)then (4) yields (for the case of no-tax free dividends):
. (6)( ) ( )0
0 , 1 1tTt i i d dT
i
PP D
Dτ τ
ΔΔ = − ⇒ = −
Expression (6) has been used in prior studies to test for tax effects inex-dividend pricing. However, (5) differs from (6) in the followingways. In (5) the cash flows are discounted back to the present whereasin (6) it is assumed that taxes and dividends are paid on the ex-day. Thetime difference between ex-day and the payment of dividends isapproximately 16 days on average and it is usually of little importancewhether the delay between ex-day and actual payment of the dividendsis accounted for. The dividend tax liability, however, may not be due forup to two years so the net present value may be significantly differentto the actual amount of tax. In particular this is the case in periods withhigh interest rates. Ddtl is different to DT so it is not possible to use a
straight inverted, divided yield or drop ratio, , to test for( )0 1
Tt tP P D−
passive or long term traders in ex-dividend pricing. The correct measure
is an adjusted drop ratio: . Since Ddtl < DT the standard( )0 1
dtlt tP P D−
drop measure given in (6) yields a downward biased estimate of τd. Forthe case of two types of dividends it is advantageous to work with thefollowing expression for empirical tests:
(7)0 , 0 2 ,
AD dtlt i i i i tP D Dγ γ εΔ = − +
The test for a passive or long term investor is:
.0 21and dγ γ τ= =
B. Active or Short Term Trader
As with passive traders there are two possible strategies for activetraders; buy or sell at the cum-day or at the ex-day.
Buying the Share Cum-Dividends:
The payoff from this strategy is derived from (1) and is given by:
(8) ( ) ( )0 0
T NT dtlt t c dE P CG D D D E Pπ τ τ= − − + + − +
Multinational Finance Journal182
Buying the Share on the Ex-Day
The expected value at the cum-day for this transaction, derived from (1),is:
( ) ( ) ( ) ( )( )0 1
s f c s
s
rT r T Trt t c tE E P e E P eπ τ − + −− ⎡= − − ⎢⎣
(9)
( ) ( )( ) ( )1
1f cr r T
tE P e E P− + − ⎤− +⎥⎦
The capital gains component for the sales price differs from (8) sincethe purchase price is not known until the ex-day but expectations aretaken at time zero. Thus the expected purchase price is discounted backfor one day using the risk adjusted rate. Using the followingapproximation
(10) ( ) ( )( ) ( ) ( )1 0 0 1
1 1f c c f f cr r T T r r Trt t t tE P e P e P E P e e
− + − − − −⎡ ⎤− + ≈ −⎣ ⎦
and setting (8) equal to (9) and collecting term yields:
(11) ( ) ( )0 1 11 f c
T NT tdlr d
t t r Tc
D D DP E P e
e
ττ
−− −
+ −− =−
Using the same notation for stock “i” as for the long term trader in (5),(11) can be rewritten as:
. (12)( ) ( )0, 1
1
1i f c
AD dtlt i d ir T
c
P D De
ττ − −Δ = −
−
The empirical model used to test for an active trader is:
(13)( ) ( )0, 2 ,1
1
1
1i f c
AD dtlt i i i tr T
P D De
γ εγ − −Δ = − +
−
and a general test for active traders is:
.1 2andc dγ τ γ τ= =
183Ex-Dividend Pricing
Notice that the exponential expression in the denominator of (13) isdifferent for different stocks since the number of days until taxes areactually paid differs across stocks. As a result of this, (13) is non-linearin the parameters.
To explore the impact of using discounted dividends and taxpayments, “Tp”, “Tc” and “Td” are set to zero in (12) for the case oftaxable dividends only:
(14) ( )( )
( )( )
0
0 ,
1 1
1 1td dT
t i i Tc c
PP D
D
τ ττ τ
Δ− −Δ = ⇔ =
− −
As for the case of passive or long term investors, (14) is equivalent tothe expression used in the literature to test for active traders. A test forthe existence of an active trader, who pays the same income tax rate onboth dividends and capital gains, is a test of whether or not the dropratio is equal to one. Equations (13) and (14) are significantly different.In particular, (13) is non-linear in the parameters whereas (14) is asimple linear t-test of the drop ratio. If there is a significant delaybetween the time of trade and payment of taxes, (14) (and (6)) ismisspecified and cannot be used to test for active (passive) traders.
C. Mixed Investor
The last type of investor is a mixed investor who has the option of beingtaxed either as a long term investor or as a short term investor. Investorswho maintain two sets of books and purchase shares on thecum-dividend day have the option of being taxed on the subsequent saleeither as a passive or active trader. If they buy the share cum-dividend,they can then choose to be taxed as an active trader by selling the stockon the ex-dividend day or as a passive trader by selling the stock at atime in the future. While the U.S. specifies a long term trade based onthe specific time period that the stock is held, the New Zealand tax codedoes not specify for how long a stock has to be held to be classified asa passive investment. Only the intent at the time of purchase matters.The implementation of such a law is obviously difficult, but one of themain arguments a trader can use to prove that a transaction should beentered into the active trading book is the length of time the stock isheld. In this paper we, therefore, assume that the stock is sold after oneday (the ex-day) if it is an active trade and held for s days (s > 1) if it is
Multinational Finance Journal184
a passive trade. Using only one day to maturity is likely to understatethe value of the option and bias the results in favour of not finding anymixed investors in the sample.
Mixed Investor Buying Cum-Dividends
The payoff at t1 to a mixed trader is therefore given by:
, (15) ( )1 1 1
max ,M P At t tπ π π=
where is the payoff if the investor chooses to be taxed as a passive1
Ptπ
or long term trader and is the payoff from an active or short term1
Atπ
strategy. The payoff at time t1 to the passive or long trading strategy isderived from (2) by taking the expectations at time t1 instead of at timet0, adjusting the number of days to tax payments, and noticing that
has already been paid and is therefore not part of the payoff at time0t
P
t1. The payoff at time t1 for the active trading strategy can be obtainedfrom (9) by making the same adjustments as for the passive tradingstrategy. The payoffs to the two types of trading strategies are:
( ) ( ) ( )1
1p f
p p
T rP T NTt t tE D D eπ − −= +
( ) ( ) ( )1 1dt f s
dt s
T r r TTd t tD e E P eτ − − − −− +
and
( ) ( ) ( ) ( )1 1 0
11 p fc f
p p
T rT rA T NTt c t t t tP P e D D eπ τ − −− −= − − − +
.( )1
1d fT rTd tD e Pτ − −− +
The active trader sells the stock at time t1 whereas the passive tradersells at time ts. Thus the payoff to the active trader is known withcertainty at time t1 but the payoff to the passive or long term tradingstrategy is a random variable. The future price, however, is discountedback at the risk adjusted rate and the expected payoff to the long termor passive trading strategy can therefore be compared to the certainpayoff from the active strategy. Substituting (16) and (17) into (15),collecting terms and adopting the convention from above that variables
185Ex-Dividend Pricing
without time subscripts are present values yields the following payoffto the mixed investor at time t1:
( ) ( )1M T NT dtlt dE D D D E Pπ τ= + − +
( ) ( ) ( )1 1 0
1max 0, c fT rt c t tP E P P P eτ − −⎡ ⎤− − − −⎣ ⎦
The payoff is the present value of after-tax dividends plus the presentvalue of the expected future sales price plus an option. The first term inthe option, is the difference between the price at t1 and the( )
1,tP E P−
present value of the sales price. In an efficient market we would expectthis term to be zero. However, if the trader has inside information, thisterm can be negative or positive. The second term in the option,
,( ) ( )1 0
1c fT rc t tP P eτ − −− −
s the capital gains tax payable if the stock is bought cum-dividend andsold ex-dividend. The tax is payable in T–1 days and is discounted at therisk free rate since it is know at time t1. In general, the second term isexpected to be negative since the share price drops when the stock goesex-dividend. The second term induces the trader to behave as a shortterm trader because doing so adds positive value to the overall trade. If,however, there are gains from insider information then the trader willhold on to the stock, i.e., behave as a long term trader.
The stochastic variable driving the option value is the share priceand the exercise price is We can further write the expected payoff
0.tP
to the mixed trader as:
( ) ( )1
M T NT dtlt dE D D D E Pπ τ= + − +
,( )0 1
1 max 0,c fT rc t te P Pτ ϖ− − ⎡ ⎤+ + −⎣ ⎦
where
( ) ( )( )11
1c f
tT rc
P E Pe
ϖτ − −= −
represents the gains from insider information.The expected value of the payoff at time t0, the cum-day, is given by:
Multinational Finance Journal186
( ) ( )0 0
M T NT dtlt t dE P D D D E Pπ τ= − + + − +
( ) ( )0 1
1 , , , , ,c fT rc f t te C r P Pτ σ ω γ− −+
here C(@) represents the option value, s is the standard deviation of thestock returns, γ is the jump parameter due to the stock going ex-dividend,(which in this case is equivalent to the drop in share price between thecum and ex-day), ω is the expected return to insider information. Thematurity of the option is very short, and is the same for all the stocks, andrf is the risk free rate of interest. An increase in the volatility will increasethe value of the option, an increase in the drop ratio will decrease boththe value of and E(P), leaving the first term unaffected. An increase
1tP
in the dividend yield, however, increases the capital loss from thedecrease in the price and so increases the tax deduction to the activetrader and the value of the option. An increase in the return from insiderinformation will decrease the value of the option because being a passive,long term, investor is more valuable. Finally, an increase in T, thenumber of days to the payment of tax, decreases the value of the optionby reducing the tax benefit, by reducing the present value of thetax-deduction from being an active investor.
The mixed investor also has the opportunity of buying the stock onthe ex-day and being taxed as a long term investor. The expected payoffas of the cum-dividend day, t0, is given by:
(20)( ) ( ) ( )0 1
s
s
rTrt t tE E P e E P eπ −−= − +
For the markets to be in equilibrium, the expected payoffs given by (19)and (20) must be the same. Therefore, the equilibrium condition for themixed trader is:
( )0
T NT dtlt dP D D D E Pτ− + + − +
( ) ( )0 1
1 , , , , ,c fT rc f t te C r P Pτ σ ω γ− −+ ( ) ( )
1
rtE P e E P−= − + ⇔
(21)
( )0 1
r T NT dtlt t dP E P e D D Dτ−− = + −
( ) ( )0 1
1 , , , , ,c fT rc f t te C r P Pτ σ ω γ− −+
187Ex-Dividend Pricing
6. The model is also relevant for other countries. For example, the U.S. allows aninvestor to be taxed as a short term investor if the stock is sold within, e.g., 6 months
As for the other cases this expression has an empirical counterpartwhich can be written as:
, (22)0 , 2 1 , ,
AD dtlt i i i i t i tP D D adOptionγ β εΔ = − + +
where( ) ( )
0 1
1, , , , ,c fT r
i t f t tadOption e C r P Pσ γ− −=
is the adjusted value of the option. The left hand side represents theexpected drop in price and the right hand side is the net present value ofthe dividends. Compared to previous studies, (20) explicitly takes intoaccount that taxes are not payable immediately. The model alsoexplicitly integrates the option generated by the specific tax laws inNew Zealand. Similar options would be available in other tax regimeswhere specific criteria determine the tax status of gains from stocktrades.
The three equilibrium conditions represented by the equations (7),(13) and (22) are nested into an empirical model that can be tested. Theempirical model is given by:
( ) ( )0 , , 2 ,1
1
1
1 cit f
AD dtlt i i t i tT r
P D De
α γγ − −Δ = + −
−
.1 , ,i t i tadOptionβ ε+ +
An intercept term was added to capture any constant effects.The predicted values of the parameters in the nested empirical model
(23) for the different traders are equivalent to the individual modelsgiven by (7), (13) and (23). These are summarized in table 1. Notice thatactive traders in New Zealand face the same income tax rate ondividends and capital gains.
In summary, in this section it was argued that in New Zealand threedifferent types of investors may dominate different stocks around theex-day. The active and passive investors are equivalent to the onestested for in the U.S. and other countries. A third type of investor is alsoidentified here, namely, a mixed investor who has the option of beingtaxed as either an active of passive investor.6 Finally, the fact that taxes
Multinational Finance Journal188
(depending on the time period).
incurred as a consequence of trading around the ex-day may not bepayable for nearly two years is explicitly taken into account. Ifincremental taxes, as a consequence of trading around the ex-day, arenot due for a while after the actual trade then, as shown above, themodels used in the literature to test the various ex-day pricing theoriesare misspecified.
Before providing the results obtained from testing (23), the data usedis discussed in the next section.
IV. Data
The data consists of daily price quotations of listed companies on theNew Zealand Stock Exchange between 1 January 1982 and 31 July 1985and the dividends on those shares for the same periods. All data entrieshave been individually checked. The database also includes adjustmentfactors for capital changes and splits and price lags, i.e., the age of theprice or when the stock was last traded. The database is compiled andmaintained by the Department of Finance and Quantitative Analysis atthe University of Otago.
To be included in the study, shares had to satisfy the followingrequirements:
TABLE 1. Parameter Predictions
γ1 γ2 β1
Passive trader 0 –τ 0Mixed trader 0 –τ +Active trader τ –τ 0
Note: The predictions for the three type of traders are based on the model
.( ) ( )0, 2 ,1
1
1
1i f cti
AD TNt i i i tr T
P D De
γ εγ − −Δ = − +
−
where r is the risk adjusted discount rate, TCit he number of days to( )0 0 1
,rt t tP P E P e−Δ = −
the payment of tax, rf is the risk free rate of return. are the present values of, ,and AD TNi t i tD D
dividends paid at the ex-day and payable in T days, respectively. AdOption is the value of thetax tax option of switching between being a active and passive investor. τ is the marginalincome tax rate.
189Ex-Dividend Pricing
Be listed in a window of –40 to +40 around the ex-day.
Be traded on the cum and ex-day.
Have at least 6 prices in total in the window –40 to –5 and +5 to +40to facilitate return calculations.
A total of 469 observations fit these criteria and form the basic sample.As is evident from table 2, the number of dividend events with only onetype of dividend payment on the same day is 406, comprising 150taxable dividends and 256 non-table dividends. In a further 63 cases,both types of dividends were paid on the same day. A total of 153different companies generated the 469 dividend events, 37 companiespaid only taxable dividends and 52 companies paid only non-taxabledividends over the period. A total of 62 companies paid both types ofdividends in the period of which 38 different companies paid twodividends on the same day. Thus, the different types of dividends arerelatively evenly spread over different companies.
Figure 2 shows a histogram of the percentage drop in prices betweenthe ex and cum-day. Clearly, there are too many cases indicating nodrop to be believable. This situation may have been caused by:
Wrong ex-day. The sources were double checked. It is possible thatsome of the dates reported are not correct but the weekly diariesfrom the stock exchange were used and these should be accurate.
There was no trade on the ex-day, making the cum-price equal to theex-price. Again, data entries were checked against the newspapersand, in general, showed a trade on the relevant days.
The volume indicated in the papers on these days was rather small.This may be a result of either a special trade where dividends are
TABLE 2. Sample Composition
Only Only non- Taxable and Totaltaxable taxable non-taxabledividends dividends dividends
Number of dividend events 150 256 63 469where the firm paid ...Number of companies that paid ... 37 52 62* 153
Note: *38 different companies paid both types of dividends on the same day.
Multinational Finance Journal190
FIGURE 2.—Price Drop Between Cum- and Ex-day
included on the ex-day or overhanging limit orders, that is, limitorders that have not being adjusted for dividends or cancelled.
Furthermore, these “outliers” were analyzed with respect to dividendyield and thin trading. If these stocks all had low dividend yield,transactions costs may prohibit trading around the ex-day to exploitprofit possibilities. As can be seen from table 3, there is no apparentdifference in dividend yield between the stocks with price changes fromthe cum-day to the ex-day and the ones for which the price did notchange. Both samples have an annual dividend yield of about 6% or 3%per dividend payment (semi-annual payments). Nor does the thin tradingmeasure show any significant difference. In summary, it is not apparentwhy prices did not change on these stocks, and it is not clear how theseobservations should be treated. The analysis below is, therefore,undertaken with and without these stocks.
In an effort to overcome data problems and to exploit the uniquefeature of the data set, where in the same firm could pay both taxableand non-taxable dividends on the same day, the analysis was performedon three data sets. The first data-set contains all 469 observations. In thesecond data set the observations with no change in the price around theex-day were dropped leaving a total of 370 observations. Finally, the
191Ex-Dividend Pricing
third set contains only stocks with a taxable and a non-taxable dividendon the same day; with 63 observations.
Table 4 contains the descriptive statistics for the different dropmeasures, ,scaled by the total dividends. If passive traders are the
0 ,t iPΔdominant traders, the drop measure should be below one for taxabledividends and one for non-taxable dividends. Consistent with a passivetrader, the drop ratio is significantly larger for non-taxable dividends;the T-test for differences in means between the taxable and non-taxabledividends are significant for all three measures. However, thecoefficient for the non-taxable dividends is significantly smaller thanone for all measures. If active traders were the dominant traders, thedrop measure should equal one for taxable dividends and be larger thanone for non-taxable dividends. This is clearly not the case. The data isconsistent with the existence of passive traders and a general tax effect.
A. Dependent Variable
The dependent variable in the empirical model (23) is calculated in twodifferent ways:
. (24)( )0 1
rt tmean adjusted P E P e−= −
In this measure, the difference between the cum and ex-dividend priceis discounted by the average daily return of the stock in window –40 toplus 40 days around the ex-day excluding plus/minus five days aroundthe ex-day. The second measure is:
TABLE 3. Analysis of “Outliers”
Stock where Pt–1 = Pt Stock where Pt–1 … Pt
Dividend yield 0.0285 0.0307Thin trading 0.8507 0.8694Number of obs. 98 371
Note: From figure 1 there may be an abnormal number of stocks where the price did notchange between the ex- and cum-days. The dividend yield is the total dividends (taxable plusnon-taxable) paid divided by the cum-price. The thin trading measure is the number of daysmeasured in percentage, the stock traded in the window –40 to +40, excluding five daysaround the dividend..
Multinational Finance Journal192
TA
BL
E 4
.D
escr
ipti
ve S
tati
stic
s
Dro
p m
easu
res
Div
iden
d yi
eld
(Sem
i ann
ual)
No
Mea
nM
arke
tT
axab
leN
on-t
axab
leV
olat
ilit
yT
hin
Insi
der
Ado
ptio
n/N
o.ad
just
men
tad
just
men
tad
just
men
tdi
vide
nds
divi
dend
str
adin
gtr
adin
gpr
ice
Obs
Tax
able
div
iden
dsM
ean μ
td
0.47
990.
5679
0.50
340.
0286
na0.
0167
0.80
636.
0639
0.15
8315
0S
td0.
9063
0.90
540.
9011
0.01
47na
0.00
790.
2215
28.1
353
0.07
06
Non
-tax
able
div
iden
ds
Mea
n μ
ntd
0.77
460.
8424
0.79
91na
0.03
010.
0168
0.89
523.
237
0.15
8325
6S
td1.
1022
1.10
091.
0972
na0.
0239
0.00
710.
1376
24.1
451
0.06
63T
-tes
ts:
μtd =
μnt
d2.
772.
582.
63na
na0.
2137
4.99
1.07
020.
0001
406
μnt
d =
13.
492.
443.
12na
nana
nana
na25
6μ
td =
16.
165.
125.
91na
nana
nana
na15
0
Bot
h ta
xabl
e an
d no
n-ta
xabl
e di
vide
nds
on th
e sa
me
day
Mea
n0.
5077
0.54
960.
5251
0.01
560.
0191
0.01
710.
8854
2.39
430.
1576
63S
td1.
4483
1.47
41.
4443
0.00
810.
0153
0.01
310.
1797
21.9
281
0.09
93
(Con
tinu
ed)
193Ex-Dividend Pricing
TA
BL
E 4
.(C
onti
nued
)
Not
e: T
he th
ree
drop
mea
sure
s ar
e de
fine
d as
: No
adju
stm
ent (
Pt–
1–P
t)/(T
otal
div
iden
ds),
Mea
n ad
just
men
t (P
t–1–
Pt/(
1+μ s
)/(t
otal
div
iden
ds)
and
mar
ket a
djus
tmen
t (P
t–1–
Pt/(
1+E
(rs)
/(to
tal d
ivid
ends
) , w
here
Pt–
1 is
the
stoc
k pr
ice
cum
-div
iden
ds, T
otal
div
iden
ds is
the
sum
of
taxa
ble
and
non-
taxa
ble
divi
dend
s. μ
s is
the
aver
age
dail
y re
turn
for t
he p
erio
d 40
day
s be
fore
and
40
days
aft
er th
e ex
-div
iden
d ex
clud
ing
5 da
ys o
n ea
ch s
ide
of t
he e
x-di
vide
nd d
ay a
nd E
(rs)
is
the
expe
cted
ret
urn
as g
ener
ated
by
the
CA
PM
. T
he d
ivid
end
yiel
d is
div
iden
d di
vide
d by
the
sto
ck p
rice
cum
-div
iden
d an
d th
e vo
lati
lity
mea
sure
is th
e st
anda
rd d
evia
tion
of
retu
rns
over
the
abo
ve w
indo
w. T
he th
in tr
adin
g m
easu
re is
def
ined
as
the
num
ber
of d
ays
the
stoc
k tr
aded
in th
e ab
ove
win
dow
div
ided
by
the
tota
l num
ber
of d
ays.
The
insi
der
trad
ing
mea
sure
is o
btai
ned
by d
isco
unti
ngth
e st
ock
pric
e 10
0 tr
adin
g da
ys f
rom
the
ex-d
ivid
end
day
by th
e ri
sk a
djus
ted
rate
, whe
re th
e ri
sk a
djus
ted
rate
is g
ener
ated
by
CA
PM
as
abov
e,th
e sh
are
pric
e on
the
ex-d
ay is
sub
trac
ted
and
the
diff
eren
ce is
div
ided
by
the
ex-p
rice
and
mul
tipl
ied
by 1
00.
Multinational Finance Journal194
7. See Bartholdy and Riding (1994) for a discussion of these issues for New Zealanddata.
(25)( ) ( )0 1
0.06fr
t tmarket adjusted P E P eβ− += −
This measure is a risk adjusted measure. The beta is calculated as atrade-by-trade beta to account for thin trading during the window –40to +40 excluding the 11 days around the ex-day.7 The risk free rate isthe 90 bank bill rate and the excess return on the market is set to 6%;see Chay et al. (1996).
B. Valuing the Option
For the right hand side of (21), the value of the option was alsocalculated in two ways. The value of the option at time t0 is:
.( ) ( )( )0 1
1 , , ,c fT rc f t tC e C r P Pτ σ γ ω− −= −
For the purpose of pricing the option it is assumed that insiderinformation is nonexistent or small enough to have no effect on themarginal price. This assumption is likely to bias downwards the valueof the option and, therefore, also bias downwards the likelihood offinding a mixed marginal investor. The standard deviation wascalculated using daily returns for 35 days prior to the ex-day windowwhere the window is defined as plus-minus 5 days around the ex-day.The time to maturity was set to one day. Again the risk free rate was the90 day bank bill rate.
In the first model, the jump parameter was ignored, and the Blackand Scholes model was used to evaluate the option. The Europeanformula was used because the valuation was being made on the cum-dayand the option was being sold on the ex-day. The option is at the money,so the strike price is the current share price in this model.
Adjusting for the jump parameter which is a feature of the trueoption is nontrivial. The problem is clearly that the jump parameter ordrop is equal to, or a function of, the drop ratio which is the variable onthe left hand side of the regression equation (23). Given the difficultiesarising from the non-linear nature of this problem, the second methodfor calculating the value of the option uses the current price adjusted fortotal dividends. The values from the two models were regressed against
195Ex-Dividend Pricing
each other, and the results indicated almost perfect correlation betweenthe two (R2= 0.99). For further empirical work, the average of the twoestimates was used as the value of the option. As can be seen from table4, the average value of the option as a percentage of the price of thestock is around 15% for all three types of dividend payments.
C. Summary
The standard procedure in the literature is to test the predictions of thevarious equilibrium conditions and from these infer the dominantmarginal investor. In these tests it is implicitly assumed that the sameinvestor type dominates the equilibrium for all stocks. In a world withshort sales and many investors this may be feasible. But in New Zealandshort sales were prohibited during the testing period. The ability toexploit profit opportunities on the sales side was limited by the numberof shares the investors held in their portfolios. Since the profits from thevarious strategies are all stochastic, risk averse investors would bereluctant to purchase unlimited amounts of a single stock and therebygive up diversification. The implication of these restrictions is that onetype of investor may not be able to dominate all stocks at all times;different types of investors may dominate different stocks. This type ofequilibrium is referred to as a separating equilibrium. The model hereis therefore estimated under the assumption of both a dominating and aseparating equilibrium.
V. Dominating Equilibrium
The results of estimating (23), under the assumptions that one type ofinvestor dominates all trades, are presented in table 5. As indicated inthe discussion of table 1, if (23) is valid then γ2 is significant and thesame is the case for all traders as this represents the tax liability ondividends. From table 5, γ2 is only significant when the outliers
are dropped; not when all the observations or those with( )0 1
0t tP P− =
two dividends are used. When γ2 is significant the value is, as predictedby the model, close to the marginal tax rate. From table 1 the predictionsfor a passive investor are that γ2 is the only significant parameter. For amixed investor to be dominating, both γ2 and β1 must be significant andfor the active traders to dominate, both γ1 and γ2 must be significant.
Multinational Finance Journal196
TA
BL
E 5
.D
omin
atin
g E
quili
briu
m
γ 1β 1
γ 2α
Obs
R2
Mea
n ad
just
ed m
etho
d:A
ll o
bser
vati
ons
–1.
9645
0.03
24–0
.219
81.
2925
469
0.16
02(1
.1)
(2.5
3)*
(0.7
1)(1
.39)
Dro
ppin
g ob
serv
atio
ns w
here
Pt–
1 =
Pt
–0.3
514
0.02
81–0
.374
0.37
3637
00.
3169
(1.4
3)(2
.00)
*(2
.31)
*(0
.65)
Mar
ket a
djus
ted
met
hod:
All
obs
erva
tion
s–1
.954
80.
0123
–0.2
908
1.72
0346
90.
1316
(1.1
)(1
.02)
(1.0
4)(1
.85)
*D
ropp
ing
obse
rvat
ions
whe
re P
t–1
= P
t–0
.340
70.
0103
–0.4
032
0.72
2437
00.
2993
(1.4
4)(0
.84)
(2.7
1)*
(1.3
4)
Tw
o di
vide
nds
on th
e sa
me
day
Mea
n ad
just
ed m
etho
d:A
ll o
bser
vati
ons
–1.0
954
0.00
482.
1515
–1.7
385
630.
5586
(0.8
2)(0
.18)
(1.0
3)(1
.6)
Dro
ppin
g ob
serv
atio
ns w
here
Pt–
1 =
Pt
–1.1
273
0.00
62.
1593
–1.0
619
530.
5759
(0.7
6)(0
.22)
(0.9
4)(0
.91)
Mar
ket a
djus
ted
met
hod:
All
obs
erva
tion
s –
1.10
18–0
.013
2.11
85–1
.274
763
0.55
77(0
.81)
(0.5
6)(0
.99)
(1.2
1)D
ropp
ing
obse
rvat
ions
whe
re P
t–1
= P
t–1
.119
5–0
.012
92.
0947
–0.5
812
530.
5728
(0.7
5)(0
.55)
(0.9
)(0
.51)
(Con
tinu
ed)
197Ex-Dividend Pricing
TA
BL
E 5
.(C
onti
nued
)
Not
e: T
he a
bove
est
imat
es a
re o
btai
ned
usin
g eq
uati
on (
23).
The
“m
ean
adju
sted
” m
etho
d us
es th
e m
ean
of th
e st
ock
retu
rn o
ver
the
win
dow
–40
to +
40 d
ays
arou
nd th
e ev
ent d
ay (e
xclu
ding
the
even
t day
); s
ee e
quat
ion
(24)
. The
“m
arke
t adj
uste
d” m
etho
d us
es a
risk
adj
uste
d m
etho
d w
ith
trad
e-to
-tra
de b
eta
calc
ulat
ed o
ver
the
sam
e w
indo
w a
nd a
n ex
pect
ed e
xces
s re
turn
on
the
mar
ket o
f 6%
and
the
90 d
ay b
ank
bill
rat
e as
the
risk
free
rat
e of
ret
urn;
see
equ
atio
n 24
. Par
enth
eses
incl
ude
the
t-va
lues
for
the
esti
mat
es. *
stat
isti
call
y si
gnif
ican
t.
Multinational Finance Journal198
In general, table 5 provides no evidence of strictly active traders; γ1
is never significant. The results also provide only weak evidence ofmixed traders in the significance of β1 for the market adjusted dropmeasure. This parameter indicates the value of the option to thoseinvestors who can choose their tax status. In this case, the significantvalues are dependent on the choice of measurement of the price drop.There is evidence of either a dominating equilibrium of mixed tradersor a separating equilibrium of mixed and passive traders.
The problem with the results in table 5 is that if a separatingequilibrium exists, the estimates are inefficient and biased. If differentinvestors dominate different stocks, the results in table 5 represent theaverage over the three different types of investors. Given that γ1, forexample, is zero for passive traders and equal to the income tax rate foractive traders, the results in table 5 would probably lie somewhere inbetween depending on the relative number of stocks dominated by eachgroup of investors. As a consequence, the variance of the estimatewould increase making the parameter insignificant. Since the parametersare different for different groups of stocks, they are no longer constantand the estimates in table 5 may, therefore, be biased as well.Considering the weak results obtained for a dominating equilibrium, aseparating equilibrium is considered next.
VI. Separating Equilibrium
The New Zealand institutional environment adds a constraint on shortselling that limits the activities of investors relative to other regimes.Selling can only occur when shares are already held and buying islimited by the available liquidity of the investor. Investors are, therefore,constrained to trade in shares which they believe will be increasing invalue. This is in contrast to the complete market situation where shortselling provides liquidity and allows a different set of decisions relatingto expected price drops. If tax effects are different as outlined above,and if investors trade with those differences in mind, then differentclasses of investors will gravitate toward shares that maximize theircomparative advantage within the constraints described. The resultcould be that different types of stocks attract different types of marginalinvestors consistently enough to generate a separating equilibrium.
From the model derived above the following characteristics of stocksthat may add value to the various classes of traders are suggested.
199Ex-Dividend Pricing
Mixed traders, who have the option of choosing their tax status withevery trade, should prefer high volatility stocks because the volatilitywill increase the value of the option. They should also prefer shareswith little insider information, or stable expected growth, since thepresence of either of these factors will reduce the value of the option.They should also prefer 'fat' traded stocks with high liquidity tomaintain the value of the option and to be able to sell easily if the activeposition is chosen. Active traders should also prefer highly liquid stocksto maximize their ability to trade without barriers. These investorsshould also prefer stocks with high dividend yields since the drop inprice provides a tax shield not available to long term investors. Passiveinvestors should value low dividend yield shares for the oppositereason. They should also value the potential for insider informationregarding high future growth.
The above discussion, derivation of the equilibrium model for theNew Zealand situation, and prior analysis by Boyd and Jagannathan(1994) suggest four variables which might separate the observed tradingactivity into three groups of traders: volatility, thin trading (liquidity),insider trading gains, and dividend yield. These four variables aredefined in the following way:
Volatility: the standard deviation of the return on the stock in thewindow of plus/minus 40 days around the ex-dividend day excludingthe five days around the ex-dividend day.
Thin trading measures: even when the shares of a company trade onboth the cum and ex-dividend days, thin trading might be a problemin pricing if there were, for example, only two trades on the relevantdates. To quantify the possible impact of this type of thin trading inthe sample, a thin trading measure was defined by dividing thenumber trades by the number of days in the same window as for thevolatility measure.
Insider gain measure: this was obtained by discounting back thestock price 100 trading days from the ex-day at the risk-adjustedrate, using the beta and market premium as calculated above. Thedifference between the actual price and the discounted price wasused as a measure of insider gain or mispricing.
Dividend yield: The standard measure of dividend divided by price.
The values of these four variables for the different types of dividendsare presented in table 4. There does not appear to be any correlation
Multinational Finance Journal200
8. For a discussion of Cluster analysis see ch. 11 in Johnson and Dean (1982).
9. See SAS/STAT User's Guide, p. 519.
between thin trading and volatility and tax status. Taxable dividendsseem to have a smaller dividend yield compared to the non-taxabledividends. That is, firms appear to pay out a higher dividend if they arenot taxed compared to the situations where they are. The insider tradingmeasure seems to be higher for the taxable dividends than the otherdividends. There is no obvious reason for this and it may be a functionof the rather weak or dubious definition of this variable.
The empirical tests for a separating equilibrium involve two steps.In the first step each stock has to be assigned to a potential marginalinvestor based on the above four variables. In the second step (23) isestimated for each group of stocks in an attempt to test for the existenceof different marginal investors.
A. Separation of stocks into groups
The main problem is how to separate the stocks into groups with respectto the above variables. An initial attempt of straight sorting based on thevarious variables was undertaken to no avail. It was therefore decidedto use cluster analysis to divide the observations into groups.
The basic idea in cluster analysis is to group the stocks in such a waythat stocks with similar characteristics regarding the above fourvariables are assigned to the same group or cluster.8 While the specifiedvariables were postulated to influence investor behavior, the strength ofinfluence and the possible interaction of the factors were unknown.Cluster analysis was, therefore, treated more or less as a black box togroup the stocks. Then regression analysis in the form of a modifiedversion of (23) was used to test the parameter restrictions for thedifferent types of marginal investors.
The number of clusters and the assignment of observations toclusters are dependent on the method used for the cluster analysis. TheSAS cluster analysis procedure provides several hierarchical methodsbased on varying distance algorithms. Because the presence of outliersor variables with large variances in the data can have a disproportionateinfluence on the clustering, the measures for the variables (volatility,insider trading, thin trading and dividend yields) were standardised tohave a mean of zero and variance of one. In the first stage, all fourvariables were used with the following methods: Average, Centroid,Complete, Flexible, McQuitty, Median, Single and Ward.9 The second
201Ex-Dividend Pricing
stage involved using the same methods with from two to four of theseparating variables since it is not clear whether all of the variables areimportant and how they interact.
For the large data set most of the methods yielded one cluster with400+ observations out of the 469 available and two clusters with veryfew stocks; equivalent to a dominating equilibrium that was rejectedabove. The “Flexible” and “Ward” options yielded a bit more spreadand clusters produced by these procedures were, therefore, used belowin the regression analysis. Separate analysis was done on the smallerdata set with both types of dividends paid on the same day. In that case,the "complete" method produced a two cluster separation that was usedfor further analysis.
From table 6, using insider trading and volatility as separatingvariables, the Flexible procedure identified two clusters. Cluster one hada lower average thin trading and volatility measure than cluster two.This suggests that cluster one represents passive or long term traders asthey do not benefit from higher liquidity or higher option values fromhigher volatility. The insider trading measure is positive for cluster oneand negative for cluster two supporting the notion that cluster onerepresents long term or passive investors whereas cluster two representsmixed or active traders. The dividend yield is essentially the same inboth clusters.
The Ward option produced three clusters, with thin trading andvolatility as the separating variables. Cluster one is consistent withactive or mixed traders based on the highest thin trading and volatilitymeasure, but low dividend yield compared to the other clusters suggestspassive or long term traders. Cluster two has the lowest thin trading andvolatility measure, suggesting passive or long term traders, but theinsider trading measure is the lowest and the dividend yield highest inthis cluster, suggesting active or mixed traders. Cluster three has thehighest insider trading measure, but the others fall in the middle of thethree groups, providing no further support. Thus, the results aresomewhat ambiguous for the Ward option.
For the shares which paid both taxable and non-taxable dividends onthe same day, the Complete option produced two clusters based oninsider trading and volatility. The results are ambiguous, with clusterone suggesting passive traders based on thin trading and volatility, butthe insider trading measure suggesting active or mixed traders. Thedividend yields are essentially the same for both clusters.
Below the various clusters are used to test for the existence of a
Multinational Finance Journal202
TA
BL
E 6
.Se
para
ting
Equ
ilibr
ium
– M
eans
of
Var
iabl
es f
or E
ach
Clu
ster
Tw
o di
vide
nds
All
obs
erva
tion
son
the
sam
e da
y
Fle
xibl
eW
ard
Com
plet
e
Clu
ster
1C
lust
er 2
Clu
ster
1C
lust
er 2
Clu
ster
3C
lust
er 1
Clu
ster
2
Num
ber
of o
bs.
183
283
134
266
6926
36T
hin
trad
ing
0.84
140.
8807
0.89
990.
8494
0.86
040.
8354
0.92
4In
side
r tr
adin
g 23
.417
7–9
.905
7–6
.974
3–0
.736
343
.761
1–1
7.69
415
.760
Div
iden
d yi
eld
0.03
0.03
040.
0187
0.03
710.
0263
0.03
410.
0359
Vol
atil
ity
0.01
470.
0177
0.02
240.
0136
0.01
850.
0138
0.01
72
Not
e: T
he s
ampl
e is
spl
it in
to c
lust
ers
usin
g cl
uste
r an
alys
is. T
he s
epar
atin
g va
riab
les
are
the
foll
owin
g: th
in tr
adin
g de
fine
d as
the
num
ber
of d
ays
the
stoc
k tr
aded
in a
69
day
win
dow
aro
und
the
cum
-day
. Ins
ider
trad
ing
is th
e di
ffer
ence
bet
wee
n th
e pr
ice
cum
-div
iden
d an
d th
e pr
ice
in 1
00 d
ays
disc
ount
ed a
t the
ris
k ad
just
ed r
ate.
Div
iden
d yi
eld
is th
e di
vide
nd d
ivid
ed b
y th
e cu
m p
rice
. Vol
atil
ity
is th
e st
anda
rd d
evia
tion
of
the
retu
rn o
n th
e st
ock
over
a 6
9 da
y w
indo
w a
roun
d th
e cu
m-d
ay. T
he f
lexi
ble
proc
edur
e us
ed th
e in
side
r tr
adin
g m
easu
re a
nd v
olat
ilit
y to
clu
ster
obse
rvat
ions
. The
War
d pr
oced
ure
used
thin
trad
ing
and
vola
tili
ty to
clu
ster
the
obse
rvat
ions
. The
com
plet
e pr
oced
ure
used
insi
der
trad
ing
and
vola
tili
ty to
clu
ster
the
obse
rvat
ions
.
203Ex-Dividend Pricing
separating equilibrium. For a separating equilibrium the parameterestimates from (23) should be significantly different from the clustersidentified by the Cluster analysis. This is the topic for the next section.
B. Regression Analysis - Tests for a Separating Equilibrium
While the cluster analysis showed that the observations could beclassified by sorting based on the specified variables in at least threedifferent ways, the results are only suggestive. To analyze each of theclusters further, the following revised version of (23) was used:
0 1 1, 2 2, 3 3,t t t tP a Z a Z a ZΔ = + +
(26)( ) ( ) ( ), 2 ,1
1,1 1, 1,2 2, 1,3 3,
1
1 cit f
AD TNi t i tT r
t t t
D DZ Z Z e
γγ γ γ − −+ +
− + +
1,1 1, , 1,2 2, , 1,3 3, , ,t i t t i t t i t i tZ adOption Z adOption Z adOptionβ β β ε+ + + +
where Zi,t = 1 for i =1, 2 and 3 if the observations belong to cluster i andzero otherwise. Notice that there is no dummy variable on the dividendtax term since all the investors face the same dividend tax rate. TheFlexible and Complete options only require two cluster dummies. If aseparating equilibrium exists, and the above cluster analysis is able toidentify the various groups of stocks associated with each marginalinvestor, then (26) should provide different estimates for the parametersacross the different clusters each representing a marginal investor.
The results from estimating (26) for the larger data set are presentedin table 7. The results for the smaller set with both taxable andnon-taxable dividends paid on the same day are discussed below. Astable 7 indicates, γ2 is significant in all runs regardless of clusteringmethods, and the value is around 40%. Based on the actual rates for theperiod, this is a reasonable figure for the average marginal tax rate ondividends. This result alone indicates an improved model compared tothe dominating equilibrium result and suggests that there is indeed ataxation effect present in ex-dividend pricing. Furthermore, as is evidentfrom table 7, γ1,i is insignificant regardless of clustering method anddrop measure used which suggests that strictly active traders were notpresent in the market.
From the Flexible clustering option in table 7, there is evidence of
Multinational Finance Journal204
TA
BL
E 7
.R
egre
ssio
n R
esul
ts f
rom
Clu
ster
Ana
lysi
s
α 1α 2
α 3
γ 1
,1γ 1
,2γ 1
,3γ 2
β 1,1
β 1,2
β 1,3
H1
H2
R2
A. C
lust
er a
naly
sis:
Fle
xibl
e op
tion
Mea
n ad
just
ed m
etho
d: A
ll o
bser
vati
ons
–0.1
213
1.59
24na
–0.2
393
–3.3
774
na–0
.426
50.
0181
0.03
94na
2.15
898.
5634
0.21
33(0
.2)
(1.7
6)na
(1.1
)(0
.97)
na(2
.97)
*(1
.07)
(2.7
2)*
na(0
.34)
(0.0
1)
Mea
n ad
just
ed m
etho
d: D
ropp
ing
obse
rvat
ions
whe
re P
t–1
= P
t
–0.0
378
0.65
17na
–0.1
704
–0.5
663
na–0
.394
10.
0363
0.03
27na
2.46
298.
5701
0.34
37(0
.06)
(0
.88)
na(0
.92)
(1.2
7)na
(2.6
8)*
(2.0
2)*
(2.0
9)*
na(0
.29)
(0.0
1)
Mar
ket a
djus
ted
met
hod:
All
obs
erva
tion
s
0.36
082.
0307
na–0
.167
8–3
.377
na–0
.449
8–0
.020
80.
0237
na1.
7433
5.04
870.
1854
(0.6
1)(2
.27)
*na
(0.9
1)(0
.96)
na(3
.32)
*(1
.24)
(1.8
6)*
na(0
.42)
(0.0
8)
Mar
ket a
djus
ted
met
hod:
Dro
ppin
g ob
serv
atio
ns w
here
Pt–
1 =
Pt
0.42
351.
0645
na–0
.102
9–0
.537
7na
–0.4
171
–0.0
023
0.01
73na
2.02
191.
6984
0.32
16(0
.67)
(1.5
3)na
(0.6
8)(1
.25)
na(3
.09)
*(0
.13)
(1.3
)na
(0.3
6)(0
.43)
(Con
tinu
ed)
205Ex-Dividend Pricing
TA
BL
E 7
.(C
onti
nued
)
B. C
lust
er a
naly
sis
- W
ard
Opt
ion
Mea
n ad
just
ed m
etho
d: A
ll o
bser
vati
ons
–1.4
621.
7736
–0.6
029
0.18
48–5
.096
2–0
.057
6–0
.400
40.
0172
0.08
5–0
.007
71.
3167
11.3
810.
2404
(2.2
6)(1
.99)
*(0
.67)
(0.7
5)(0
.83)
(0.2
5)(3
.0)*
(0.8
4)(3
.24)
*(0
.3)
(0.7
3)0.
0
Mea
n ad
just
ed m
etho
d: D
ropp
ing
obse
rvat
ions
whe
re P
t–1
= P
t
–0.8
672
0.69
88–0
.458
60.
2368
–0.7
026
–0.0
496
–0.3
657
0.01
530.
0571
–0.0
009
2.57
984.
2381
0.33
82(1
.08)
(0.9
5)(0
.37)
(0.9
5)(1
.28)
(0.1
8)(2
.53)
*(0
.68)
(1.9
2)*
(0.0
2)(0
.46)
(0.2
4)
Mar
ket a
djus
ted
met
hod:
All
obs
erva
tion
s
–1.1
022
2.07
09–0
.004
70.
2537
–5.0
761
–0.0
728
–0.4
495
–0.0
026
0.07
–0.0
425
2.58
5311
.469
30.
229
(1.9
2)*
(2.3
1)*
0.0
(1.3
2)(0
.83)
(0.3
9)(3
.83)
*(0
.16)
(2.5
6)*
(2.1
6)*
(0.4
6)(0
.0)
Mar
ket a
djus
ted
met
hod:
Dro
ppin
g ob
serv
atio
ns w
here
Pt–
1 =
Pt
–0.5
326
0.93
380.
1123
0.29
92–0
.689
7–0
.069
8–0
.398
8–0
.004
40.
0451
–0.0
344
4.13
993.
7091
0.33
41(0
.74)
(1.3
3)(0
.09)
(1.5
4)(1
.28)
(0.3
2)(3
.1)*
(0.2
4)(1
.49)
(1.1
8)(0
.25)
(0.2
9)
Not
e: T
he a
bove
est
imat
es a
re o
btai
ned
usin
g eq
uati
on (
26).
The
“m
ean
adju
sted
” m
etho
d us
es th
e m
ean
of th
e st
ock
retu
rn o
ver
the
win
dow
–40
to +
40 d
ays
arou
nd th
e ev
ent d
ay (e
xclu
ding
the
even
t day
); s
ee e
quat
ion
(24)
. The
“m
arke
t adj
uste
d” m
etho
d us
es a
risk
adj
uste
d m
etho
d w
ith
trad
e-to
-tra
de b
eta
calc
ulat
ed o
ver
the
sam
e w
indo
w a
nd a
n ex
pect
ed e
xces
s re
turn
on
the
mar
ket o
f 6%
and
the
90 d
ay b
ank
bill
rat
e as
the
risk
free
rate
of r
etur
n; s
ee e
quat
ion
24. P
aren
thes
es in
clud
e th
e t-
valu
es fo
r the
est
imat
es. H
1: is
the
hypo
thes
is γ 1
,1=γ 1
,2=
γ 1,3=
0 a
nd H
2: is
the
hypo
thes
isβ 1
,1=
β1,
2= β
1,3=
0. T
he n
umbe
rs u
nder
H1
and
H2
are
chi-
squa
red
stat
isti
cs f
or t
esti
ng t
he n
ull
hypo
thes
es;
the
corr
espo
ndin
g nu
mbe
rs i
npa
rent
hese
s ar
e th
eir
sign
ific
ance
leve
l. P
aren
thes
es *
Sta
tist
ical
ly s
igni
fica
nt.
Multinational Finance Journal206
mixed traders since the option value β2 is mostly significant. The lackof significance of and γ1,i and β1 (in most cases) suggests that cluster onerepresents long term investors. There is evidence in favor of mixed andpassive traders generating a separating equilibrium and no evidence ofthe existence of strictly active traders when the Flexible clusteringoption is used.
For the Ward clustering option γ2 is again consistently significantlending support to the overall model, and is consistently insignificant.This supports the finding from the Flexible clustering option that strictlyactive traders were not present. β1 is consistently insignificantsuggesting that cluster one is dominated by passive or long term traders.β2 is significant in most cases suggesting that cluster two is dominatedby mixed traders. Finally, cluster three β3 is significant with the wrongsign when the market adjusted drop measure is used for the largesample, a result somewhat difficult to interpret. Overall, the Wardoption also suggests the presence of two types of investors namelypassive or long term traders and mixed traders and provides no evidenceof strictly active traders. This is consistent with table 7 and the Flexibleoption.
Finally, the existence of the active and mixed traders was tested forby linear restrictions on the parameters. If none of the clusters representactive traders, then none of the γ1,i are significant. Thus the followinghypothesis can be tested:
H1: no active traders, or γ1,1 = γ1,2 = γ1,3 = 0
The evidence from table 7 shows that this hypothesis cannot be rejectedregardless of the clustering method and drop measure used. Thissupports the results from the above. If all active traders are able to keeptwo books, they would prefer to do so as this represents a tax option tothem. Thus, an “active” trading strategy is carried out by mixed traders.The next hypothesis, therefore, is to test whether or not mixed investorsare present in the market. If no mixed investors are present, none of theβs are significant. This can be tested in the following way:
H2 : no mixed trader hypothesis, or β1,1 = β1,2 = β1,3 = 0
From table 7 there is evidence that this hypothesis is rejected regardlessof the Cluster method used. Thus, it appears that mixed investors arepresent in the market, again supporting the results above.
207Ex-Dividend Pricing
TA
BL
E 8
.St
ocks
wit
h T
wo
Div
iden
ds o
n th
e Sa
me
day
α 1α 2
α 3
γ 1
,1γ 1
,2γ 1
,3γ 2
β 1,1
β 1,2
β 1,3
H1
H2
R2
Clu
ster
ana
lysi
s: C
om o
ptio
n, tw
o cl
uste
rs g
ener
ated
Mea
n ad
just
ed m
etho
d: A
ll o
bser
vati
ons
–2.1
404
1.23
5na
0.22
23–4
.282
1na
0.04
61–0
.040
40.
0638
na1.
8378
7.71
710.
6553
–1.5
5(0
.85)
na(1
.07)
(0.7
1)na
(0.0
9)(0
.67)
(2.6
9)*
na(0
.4)
(0.0
2)
Mea
n ad
just
ed m
etho
d: D
ropp
ing
obse
rvat
ions
whe
re P
t–1
= P
t
–0.9
369
2.73
23na
0.33
94–9
.926
1na
–0.3
739
–0.0
140.
0617
na11
.950
87.
8505
0.72
31(1
.11)
(1.8
6)na
(3.4
3)*
(0.4
)na
(2.2
1)*
(0.3
5)(2
.78)
*na
(0.0
)(0
.02)
Mar
ket a
djus
ted
met
hod:
All
obs
erva
tion
s
–1.8
764
1.46
2na
0.12
86–3
.577
7na
0.19
98–0
.039
50.
0372
na1.
0191
3.47
140.
6324
(1.3
8)(0
.98)
na(0
.52)
–0.7
4na
(0.3
7)(0
.65)
(1.7
3)na
(0.6
)(0
.18)
Mar
ket a
djus
ted
met
hod:
Dro
ppin
g ob
serv
atio
ns w
here
Pt–
1 =
Pt
–0.7
472
2.93
69na
0.26
35–6
.785
5na
–0.2
574
–0.0
136
0.03
59na
4.62
843.
4265
0.69
72(0
.96)
(1.8
7)na
(2.0
7)*
(0.4
8)na
(1.3
2)(0
.34)
(1.8
2)*
na(0
.09)
(0.1
8)
Not
es:
see
tabl
e 7
for
expl
anat
ions
.
Multinational Finance Journal208
The general conclusion is that marginal investors were never strictlyactive traders, but that both mixed and strictly passive investors weremarginal traders for different types of shares in the New Zealand sharemarket in the period covered by the research.
Table 8 reports the results for the shares with two dividends on thesame day. Due to the small sample size, none of the parameters areestimated with high confidence. The only set of results which produceda significant value for γ2 was the average adjusted price drop for thesample with dropped observations. In that case the capital gains taxvalue was significant for cluster one while the option value wassignificant for cluster two. This suggests that active investors and mixedtraders were attracted to these shares while passive investors were not.One possible rationale for this behavior is that the price drop would beexpected to be larger for these shares than for those paying only taxabledividends. The price drop will be a tax deductible loss for activeinvestors, but a non-deductible loss to passive investors.
VII. Summary and Conclusions
This paper has developed a model of the pricing behavior for sharesaround the ex-dividend day based on the unique institutionalcharacteristics of the New Zealand tax code between 1982 and 1985.The restrictions on short selling in place in New Zealand over the timeperiod reinforced the possibility that different segments of the marketwere attractive to different types of traders as classified under the taxregime. The empirical results indicate that no single type of investoremerges as the marginal trader dominating all stocks; a dominatingequilibrium. There is evidence, however, for a separating equilibriumin which passive, long term, investors emerge as the marginal traders inone segment while mixed traders, who can determine their tax status atthe time of a share purchase, are the marginal traders in others.
References
Barclay, M.J. 1987. Dividends, taxes, and common stock prices: Theex-dividend day behavior of common stock prices before the income tax.Journal of Financial Economics 19: 31-44.
Bartholdy, J. and Brown, K. 1999. Ex-dividend day pricing in New Zealand.Accounting and Finance 39(2): 111-130.
209Ex-Dividend Pricing
Bartholdy, J. and Riding, A. 1994. Thin trading and the estimation of betas onthe New Zealand share market: The efficiency of alternative techniques.Journal of Financial Research 17(2): 241-254.
Booth, L.D. and Johnston, D.J. 1984. The ex-dividend day behavior ofCanadian stock prices: Tax changes and clientele effects. Journal ofFinance 39(2): 457-476.
Boyd, J.H. and Jagannathan, R. 1994. Ex-dividend price behavior of commonstocks. Review of Financial Studies 7(4): 711-714.
Brown, P. and Walter, T. 1986. Ex-dividend day behavior of Australian shareprices. Australian Journal of Management 11(2): 140-152.
Chay, J. B.; Marsden, A.; and Stubbs, R. 1995. Investment returns in the NewZealand market: 1931-1994. New Zealand Investment Analyst: 19-27.
Chui, A.; Strong, N.; and Cadle, J. 1992. The empirical significance of taxeffects on the valuation of dividends: The UK evidence. Journal of BusinessFinance and Accounting 19(4): 515-532.
Eades, K. M.; Hess, P. J.; and. Kim, E. H. 1984. On interpreting security returnsduring the ex-dividend period. Journal of Financial Economics 13: 3-34.
Elton, E.J. and Gruber, M.J. 1970. Marginal stockholder tax rates and theclientele effect. Review of Economics and Statistics 52: 68-74.
Frank, M. and Jagannathan, R. 1998. Why do stock prices drop by less than thevalue of the dividend? Evidence from a country without taxes. Journal ofFinancial Economics (February): 161-188.
Hearth, D. and Rimbey, J.N. 1993. The dividend-clientele controversy and thetax reform act of 1986. Quarterly Journal of Business and Economics32(1): 68-81.
Heath, D.C. and Jarrow, R.A. 1988. Ex-dividend stock price behavior andarbitrage opportunities. Journal of Business 61(1): 95-108.
Johnson, R.A. and Wichern, D.W. 1982. Applied Multivariate StatisticalAnalysis. Prentice Hall.
Kalay, A. 1982. The ex-dividend day behavior of stock prices: Are-examination of the clientele effect. Journal of Finance 37(4): 1059-1070.
Koski, J.L. and Scruggs, J. 1998. Who trades around the ex-dividend day?Evidence from NYSE audit file data. Financial Management 27(3): 58-72.
Lakonishok, J. and Vermaelen, T. 1983. Tax reform and ex-dividend daybehavior. Journal of Finance 38(4): 1157-1179.
Lasfer, M.A. 1995. Ex-dividend behavior: Tax or short-term trading effects.Journal of Finance 50: 875-897.
Michaely, R. 1991. Ex-dividend day stock price behavior: The case of the 1986tax reform act. Journal of Finance 46(3): 845-859.
Michaely, R. and Murgia, M. 1995. The effect of tax heterogeneity on pricesand volume around the ex-dividend day: Evidence from the Milan stockexchange. Review of Financial Studies 8: 369-399.
Robin, A. 1991. The impact of the 1986 tax reform act on ex-dividend dayreturns. Financial Management 20(1): pp. 60-70.
SAS/STAT User's Guide, Volume 1, Version 6, SAS Institute.
Recommended